Wenbo Hu

Wenbo Hu, Yifan Xu, Yi Li et al. · harvard

Vision Language Models (VLMs), which extend Large Language Models (LLM) by incorporating visual understanding capability, have demonstrated significant advancements in addressing open-ended visual question-answering (VQA) tasks. However, these models cannot accurately interpret images infused with text, a common occurrence in real-world scenarios. Standard procedures for extracting information from images often involve learning a fixed set of query embeddings. These embeddings are designed to encapsulate image contexts and are later used as soft prompt inputs in LLMs. Yet, this process is limited to the token count, potentially curtailing the recognition of scenes with text-rich context. To improve upon them, the present study introduces BLIVA: an augmented version of InstructBLIP with Visual Assistant. BLIVA incorporates the query embeddings from InstructBLIP and also directly projects encoded patch embeddings into the LLM, a technique inspired by LLaVA. This approach assists the model to capture intricate details potentially missed during the query decoding process. Empirical evidence demonstrates that our model, BLIVA, significantly enhances performance in processing text-rich VQA benchmarks (up to 17.76% in OCR-VQA benchmark) and in undertaking general (not particularly text-rich) VQA benchmarks (up to 7.9% in Visual Spatial Reasoning benchmark), and achieved 17.72% overall improvement in a comprehensive multimodal LLM benchmark (MME), comparing to our baseline InstructBLIP. BLIVA demonstrates significant capability in decoding real-world images, irrespective of text presence. To demonstrate the broad industry applications enabled by BLIVA, we evaluate the model using a new dataset comprising YouTube thumbnails paired with question-answer sets across 11 diverse categories. Our code and models are freely accessible at https://github.com/mlpc-ucsd/BLIVA.

Wenbo Hu, Xiangjun Gao, Xiaoyu Li et al.

Estimating video depth in open-world scenarios is challenging due to the diversity of videos in appearance, content motion, camera movement, and length. We present DepthCrafter, an innovative method for generating temporally consistent long depth sequences with intricate details for open-world videos, without requiring any supplementary information such as camera poses or optical flow. The generalization ability to open-world videos is achieved by training the video-to-depth model from a pre-trained image-to-video diffusion model, through our meticulously designed three-stage training strategy. Our training approach enables the model to generate depth sequences with variable lengths at one time, up to 110 frames, and harvest both precise depth details and rich content diversity from realistic and synthetic datasets. We also propose an inference strategy that can process extremely long videos through segment-wise estimation and seamless stitching. Comprehensive evaluations on multiple datasets reveal that DepthCrafter achieves state-of-the-art performance in open-world video depth estimation under zero-shot settings. Furthermore, DepthCrafter facilitates various downstream applications, including depth-based visual effects and conditional video generation.

Jingli Lin, Runsen Xu, Shaohao Zhu et al.

Spatial understanding over continuous visual input is crucial for MLLMs to evolve into general-purpose assistants in physical environments. Yet there is still no comprehensive benchmark that holistically assesses the progress toward this goal. In this work, we introduce MMSI-Video-Bench, a fully human-annotated benchmark for video-based spatial intelligence in MLLMs. It operationalizes a four-level framework, Perception, Planning, Prediction, and Cross-Video Reasoning, through 1,106 questions grounded in 1,278 clips from 25 datasets and in-house videos. Each item is carefully designed and reviewed by 3DV experts with explanatory rationales to ensure precise, unambiguous grounding. Leveraging its diverse data sources and holistic task coverage, MMSI-Video-Bench also supports three domain-oriented sub-benchmarks (Indoor Scene Perception Bench, Robot Bench and Grounding Bench) for targeted capability assessment. We evaluate 25 strong open-source and proprietary MLLMs, revealing a striking human--AI gap: many models perform near chance, and the best reasoning model lags humans by nearly 60%. We further find that spatially fine-tuned models still fail to generalize effectively on our benchmark. Fine-grained error analysis exposes systematic failures in geometric reasoning, motion grounding, long-horizon prediction, and cross-video correspondence. We also show that typical frame-sampling strategies transfer poorly to our reasoning-intensive benchmark, and that neither 3D spatial cues nor chain-of-thought prompting yields meaningful gains. We expect our benchmark to establish a solid testbed for advancing video-based spatial intelligence.

Wangbo Yu, Jinbo Xing, Li Yuan et al.

Despite recent advancements in neural 3D reconstruction, the dependence on dense multi-view captures restricts their broader applicability. In this work, we propose \textbf{ViewCrafter}, a novel method for synthesizing high-fidelity novel views of generic scenes from single or sparse images with the prior of video diffusion model. Our method takes advantage of the powerful generation capabilities of video diffusion model and the coarse 3D clues offered by point-based representation to generate high-quality video frames with precise camera pose control. To further enlarge the generation range of novel views, we tailored an iterative view synthesis strategy together with a camera trajectory planning algorithm to progressively extend the 3D clues and the areas covered by the novel views. With ViewCrafter, we can facilitate various applications, such as immersive experiences with real-time rendering by efficiently optimizing a 3D-GS representation using the reconstructed 3D points and the generated novel views, and scene-level text-to-3D generation for more imaginative content creation. Extensive experiments on diverse datasets demonstrate the strong generalization capability and superior performance of our method in synthesizing high-fidelity and consistent novel views.

Runzhe Zhu, Ling Yin, Mingze Yang et al.

Cross-view image matching aims to match images of the same target scene acquired from different platforms. With the rapid development of drone technology, cross-view matching by neural network models has been a widely accepted choice for drone position or navigation. However, existing public datasets do not include images obtained by drones at different heights, and the types of scenes are relatively homogeneous, which yields issues in assessing a model's capability to adapt to complex and changing scenes. In this end, we present a new cross-view dataset called SUES-200 to address these issues. SUES-200 contains 24120 images acquired by the drone at four different heights and corresponding satellite view images of the same target scene. To the best of our knowledge, SUES-200 is the first public dataset that considers the differences generated in aerial photography captured by drones flying at different heights. In addition, we developed an evaluation for efficient training, testing and evaluation of cross-view matching models, under which we comprehensively analyze the performance of nine architectures. Then, we propose a robust baseline model for use with SUES-200. Experimental results show that SUES-200 can help the model to learn highly discriminative features of the height of the drone.

Sijie Zhao, Wenbo Hu, Xiaodong Cun et al.

This paper presents a novel framework for converting 2D videos to immersive stereoscopic 3D, addressing the growing demand for 3D content in immersive experience. Leveraging foundation models as priors, our approach overcomes the limitations of traditional methods and boosts the performance to ensure the high-fidelity generation required by the display devices. The proposed system consists of two main steps: depth-based video splatting for warping and extracting occlusion mask, and stereo video inpainting. We utilize pre-trained stable video diffusion as the backbone and introduce a fine-tuning protocol for the stereo video inpainting task. To handle input video with varying lengths and resolutions, we explore auto-regressive strategies and tiled processing. Finally, a sophisticated data processing pipeline has been developed to reconstruct a large-scale and high-quality dataset to support our training. Our framework demonstrates significant improvements in 2D-to-3D video conversion, offering a practical solution for creating immersive content for 3D devices like Apple Vision Pro and 3D displays. In summary, this work contributes to the field by presenting an effective method for generating high-quality stereoscopic videos from monocular input, potentially transforming how we experience digital media.

Guande He, Peng Cui, Jianfei Chen et al.

Despite the significant progress made in practical applications of aligned language models (LMs), they tend to be overconfident in output answers compared to the corresponding pre-trained LMs. In this work, we systematically evaluate the impact of the alignment process on logit-based uncertainty calibration of LMs under the multiple-choice setting. We first conduct a thoughtful empirical study on how aligned LMs differ in calibration from their pre-trained counterparts. Experimental results reveal that there are two distinct uncertainties in LMs under the multiple-choice setting, which are responsible for the answer decision and the format preference of the LMs, respectively. Then, we investigate the role of these two uncertainties on aligned LM's calibration through fine-tuning in simple synthetic alignment schemes and conclude that one reason for aligned LMs' overconfidence is the conflation of these two types of uncertainty. Furthermore, we examine the utility of common post-hoc calibration methods for aligned LMs and propose an easy-to-implement and sample-efficient method to calibrate aligned LMs. We hope our findings could provide insights into the design of more reliable alignment processes for LMs.

Wenbo Hu, Yuling Wang, Lin Ma et al.

Despite the tremendous progress in neural radiance fields (NeRF), we still face a dilemma of the trade-off between quality and efficiency, e.g., MipNeRF presents fine-detailed and anti-aliased renderings but takes days for training, while Instant-ngp can accomplish the reconstruction in a few minutes but suffers from blurring or aliasing when rendering at various distances or resolutions due to ignoring the sampling area. To this end, we propose a novel Tri-Mip encoding that enables both instant reconstruction and anti-aliased high-fidelity rendering for neural radiance fields. The key is to factorize the pre-filtered 3D feature spaces in three orthogonal mipmaps. In this way, we can efficiently perform 3D area sampling by taking advantage of 2D pre-filtered feature maps, which significantly elevates the rendering quality without sacrificing efficiency. To cope with the novel Tri-Mip representation, we propose a cone-casting rendering technique to efficiently sample anti-aliased 3D features with the Tri-Mip encoding considering both pixel imaging and observing distance. Extensive experiments on both synthetic and real-world datasets demonstrate our method achieves state-of-the-art rendering quality and reconstruction speed while maintaining a compact representation that reduces 25% model size compared against Instant-ngp.

Wenbo Hu, Xin Sun, Qiang liu et al.

Post-click conversion rate (CVR) is a reliable indicator of online customers' preferences, making it crucial for developing recommender systems. A major challenge in predicting CVR is severe selection bias, arising from users' inherent self-selection behavior and the system's item selection process. To mitigate this issue, the inverse propensity score (IPS) is employed to weight the prediction error of each observed instance. However, current propensity score estimations are unreliable due to the lack of a quality measure. To address this, we evaluate the quality of propensity scores from the perspective of uncertainty calibration, proposing the use of Expected Calibration Error (ECE) as a measure of propensity-score quality, which quantifies the extent to which predicted probabilities are overconfident by assessing the difference between predicted probabilities and actual observed frequencies. Miscalibrated propensity scores can lead to distorted IPS weights, thereby compromising the debiasing process in CVR prediction. In this paper, we introduce a model-agnostic calibration framework for propensity-based debiasing of CVR predictions. Theoretical analysis on bias and generalization bounds demonstrates the superiority of calibrated propensity estimates over uncalibrated ones. Experiments conducted on the Coat, Yahoo and KuaiRand datasets show improved uncertainty calibration, as evidenced by lower ECE values, leading to enhanced CVR prediction outcomes.

Wangbo Yu, Li Yuan, Yan-Pei Cao et al.

Recent advances in diffusion models have enabled 3D generation from a single image. However, current methods often produce suboptimal results for novel views, with blurred textures and deviations from the reference image, limiting their practical applications. In this paper, we introduce HiFi-123, a method designed for high-fidelity and multi-view consistent 3D generation. Our contributions are twofold: First, we propose a Reference-Guided Novel View Enhancement (RGNV) technique that significantly improves the fidelity of diffusion-based zero-shot novel view synthesis methods. Second, capitalizing on the RGNV, we present a novel Reference-Guided State Distillation (RGSD) loss. When incorporated into the optimization-based image-to-3D pipeline, our method significantly improves 3D generation quality, achieving state-of-the-art performance. Comprehensive evaluations demonstrate the effectiveness of our approach over existing methods, both qualitatively and quantitatively. Video results are available on the project page.

Bangbang Yang, Wenqi Dong, Lin Ma et al.

Diffusion-based methods have achieved prominent success in generating 2D media. However, accomplishing similar proficiencies for scene-level mesh texturing in 3D spatial applications, e.g., XR/VR, remains constrained, primarily due to the intricate nature of 3D geometry and the necessity for immersive free-viewpoint rendering. In this paper, we propose a novel indoor scene texturing framework, which delivers text-driven texture generation with enchanting details and authentic spatial coherence. The key insight is to first imagine a stylized 360° panoramic texture from the central viewpoint of the scene, and then propagate it to the rest areas with inpainting and imitating techniques. To ensure meaningful and aligned textures to the scene, we develop a novel coarse-to-fine panoramic texture generation approach with dual texture alignment, which both considers the geometry and texture cues of the captured scenes. To survive from cluttered geometries during texture propagation, we design a separated strategy, which conducts texture inpainting in confidential regions and then learns an implicit imitating network to synthesize textures in occluded and tiny structural areas. Extensive experiments and the immersive VR application on real-world indoor scenes demonstrate the high quality of the generated textures and the engaging experience on VR headsets. Project webpage: https://ybbbbt.com/publication/dreamspace

Jiaxin Gao, Wenbo Hu, Dongxiao Zhang et al.

Electrical energy is essential in today's society. Accurate electrical load forecasting is beneficial for better scheduling of electricity generation and saving electrical energy. In this paper, we propose theory-guided deep-learning load forecasting 2.0 (TgDLF2.0) to solve this issue, which is an improved version of the theory-guided deep-learning framework for load forecasting via ensemble long short-term memory (TgDLF). TgDLF2.0 introduces the deep-learning model Transformer and transfer learning on the basis of dividing the electrical load into dimensionless trends and local fluctuations, which realizes the utilization of domain knowledge, captures the long-term dependency of the load series, and is more appropriate for realistic scenarios with scarce samples. Cross-validation experiments on different districts show that TgDLF2.0 is approximately 16% more accurate than TgDLF and saves more than half of the training time. TgDLF2.0 with 50% weather noise has the same accuracy as TgDLF without noise, which proves its robustness. We also preliminarily mine the interpretability of Transformer in TgDLF2.0, which may provide future potential for better theory guidance. Furthermore, experiments demonstrate that transfer learning can accelerate convergence of the model in half the number of training epochs and achieve better performance.

Ruijie Zhu, Jiahao Lu, Wenbo Hu et al.

We introduce MotionCrafter, a video diffusion-based framework that jointly reconstructs 4D geometry and estimates dense motion from a monocular video. The core of our method is a novel joint representation of dense 3D point maps and 3D scene flows in a shared coordinate system, and a novel 4D VAE to effectively learn this representation. Unlike prior work that forces the 3D value and latents to align strictly with RGB VAE latents-despite their fundamentally different distributions-we show that such alignment is unnecessary and leads to suboptimal performance. Instead, we introduce a new data normalization and VAE training strategy that better transfers diffusion priors and greatly improves reconstruction quality. Extensive experiments across multiple datasets demonstrate that MotionCrafter achieves state-of-the-art performance in both geometry reconstruction and dense scene flow estimation, delivering 38.64% and 25.0% improvements in geometry and motion reconstruction, respectively, all without any post-optimization. Project page: https://ruijiezhu94.github.io/MotionCrafter_Page

Wenbo Hu, Hongjian Zhan, Xinchen Ma et al.

In the field of historical manuscript research, scholars frequently encounter novel symbols in ancient texts, investing considerable effort in their identification and documentation. Although existing object detection methods achieve impressive performance on known categories, they struggle to recognize novel symbols without retraining. To address this limitation, we propose a Visually Guided Text Spotting (VGTS) approach that accurately spots novel characters using just one annotated support sample. The core of VGTS is a spatial alignment module consisting of a Dual Spatial Attention (DSA) block and a Geometric Matching (GM) block. The DSA block aims to identify, focus on, and learn discriminative spatial regions in the support and query images, mimicking the human visual spotting process. It first refines the support image by analyzing inter-channel relationships to identify critical areas, and then refines the query image by focusing on informative key points. The GM block, on the other hand, establishes the spatial correspondence between the two images, enabling accurate localization of the target character in the query image. To tackle the example imbalance problem in low-resource spotting tasks, we develop a novel torus loss function that enhances the discriminative power of the embedding space for distance metric learning. To further validate our approach, we introduce a new dataset featuring ancient Dongba hieroglyphics (DBH) associated with the Naxi minority of China. Extensive experiments on the DBH dataset and other public datasets, including EGY, VML-HD, TKH, and NC, show that VGTS consistently surpasses state-of-the-art methods. The proposed framework exhibits great potential for application in historical manuscript text spotting, enabling scholars to efficiently identify and document novel symbols with minimal annotation effort.

Sixiao Zheng, Minghao Yin, Wenbo Hu et al.

Video world models aim to simulate dynamic, real-world environments, yet existing methods struggle to provide unified and precise control over camera and multi-object motion, as videos inherently operate dynamics in the projected 2D image plane. To bridge this gap, we introduce VerseCrafter, a 4D-aware video world model that enables explicit and coherent control over both camera and object dynamics within a unified 4D geometric world state. Our approach is centered on a novel 4D Geometric Control representation, which encodes the world state through a static background point cloud and per-object 3D Gaussian trajectories. This representation captures not only an object's path but also its probabilistic 3D occupancy over time, offering a flexible, category-agnostic alternative to rigid bounding boxes or parametric models. These 4D controls are rendered into conditioning signals for a pretrained video diffusion model, enabling the generation of high-fidelity, view-consistent videos that precisely adhere to the specified dynamics. Unfortunately, another major challenge lies in the scarcity of large-scale training data with explicit 4D annotations. We address this by developing an automatic data engine that extracts the required 4D controls from in-the-wild videos, allowing us to train our model on a massive and diverse dataset.

Jiahao Lu, Jiayi Xu, Wenbo Hu et al.

Estimating the 3D trajectory of every pixel from a monocular video is crucial and promising for a comprehensive understanding of the 3D dynamics of videos. Recent monocular 3D tracking works demonstrate impressive performance, but are limited to either tracking sparse points on the first frame or a slow optimization-based framework for dense tracking. In this paper, we propose a feedforward model, called Track4World, enabling an efficient holistic 3D tracking of every pixel in the world-centric coordinate system. Built on the global 3D scene representation encoded by a VGGT-style ViT, Track4World applies a novel 3D correlation scheme to simultaneously estimate the pixel-wise 2D and 3D dense flow between arbitrary frame pairs. The estimated scene flow, along with the reconstructed 3D geometry, enables subsequent efficient 3D tracking of every pixel of this video. Extensive experiments on multiple benchmarks demonstrate that our approach consistently outperforms existing methods in 2D/3D flow estimation and 3D tracking, highlighting its robustness and scalability for real-world 4D reconstruction tasks.

Wenxuan Huang, Shuang Chen, Zheyong Xie et al.

Unified multimodal understanding and generation models recently have achieve significant improvement in image generation capability, yet a large gap remains in instruction following and detail preservation compared to systems that tightly couple comprehension with generation such as GPT-4o. Motivated by recent advances in interleaving reasoning, we explore whether such reasoning can further improve Text-to-Image (T2I) generation. We introduce Interleaving Reasoning Generation (IRG), a framework that alternates between text-based thinking and image synthesis: the model first produces a text-based thinking to guide an initial image, then reflects on the result to refine fine-grained details, visual quality, and aesthetics while preserving semantics. To train IRG effectively, we propose Interleaving Reasoning Generation Learning (IRGL), which targets two sub-goals: (1) strengthening the initial think-and-generate stage to establish core content and base quality, and (2) enabling high-quality textual reflection and faithful implementation of those refinements in a subsequent image. We curate IRGL-300K, a dataset organized into six decomposed learning modes that jointly cover learning text-based thinking, and full thinking-image trajectories. Starting from a unified foundation model that natively emits interleaved text-image outputs, our two-stage training first builds robust thinking and reflection, then efficiently tunes the IRG pipeline in the full thinking-image trajectory data. Extensive experiments show SoTA performance, yielding absolute gains of 5-10 points on GenEval, WISE, TIIF, GenAI-Bench, and OneIG-EN, alongside substantial improvements in visual quality and fine-grained fidelity. The code, model weights and datasets will be released in: https://github.com/Osilly/Interleaving-Reasoning-Generation .

Zijun Chen, Wenbo Hu, Guande He et al.

Multimodal large language models (MLLMs) combine visual and textual data for tasks such as image captioning and visual question answering. Proper uncertainty calibration is crucial, yet challenging, for reliable use in areas like healthcare and autonomous driving. This paper investigates representative MLLMs, focusing on their calibration across various scenarios, including before and after visual fine-tuning, as well as before and after multimodal training of the base LLMs. We observed miscalibration in their performance, and at the same time, no significant differences in calibration across these scenarios. We also highlight how uncertainty differs between text and images and how their integration affects overall uncertainty. To better understand MLLMs' miscalibration and their ability to self-assess uncertainty, we construct the IDK (I don't know) dataset, which is key to evaluating how they handle unknowns. Our findings reveal that MLLMs tend to give answers rather than admit uncertainty, but this self-assessment improves with proper prompt adjustments. Finally, to calibrate MLLMs and enhance model reliability, we propose techniques such as temperature scaling and iterative prompt optimization. Our results provide insights into improving MLLMs for effective and responsible deployment in multimodal applications. Code and IDK dataset: https://github.com/hfutml/Calibration-MLLM.

Wenbo Hu, Xin Chen, Yan Gao-Tian et al.

Group Relative Policy Optimization (GRPO) has emerged as the de facto Reinforcement Learning (RL) objective driving recent advancements in Multimodal Large Language Models. However, extending this success to open-source multimodal generalist models remains heavily constrained by two primary challenges: the extreme variance in reward topologies across diverse visual tasks, and the inherent difficulty of balancing fine-grained perception with multi-step reasoning capabilities. To address these issues, we introduce Gaussian GRPO (G$^2$RPO), a novel RL training objective that replaces standard linear scaling with non-linear distributional matching. By mathematically forcing the advantage distribution of any given task to strictly converge to a standard normal distribution, $\mathcal{N}(0,1)$, G$^2$RPO theoretically ensures inter-task gradient equity, mitigates vulnerabilities to heavy-tail outliers, and offers symmetric update for positive and negative rewards. Leveraging the enhanced training stability provided by G$^2$RPO, we introduce two task-level shaping mechanisms to seamlessly balance perception and reasoning. First, response length shaping dynamically elicits extended reasoning chains for complex queries while enforce direct outputs to bolster visual grounding. Second, entropy shaping tightly bounds the model's exploration zone, effectively preventing both entropy collapse and entropy explosion. Integrating these methodologies, we present OpenVLThinkerV2, a highly robust, general-purpose multimodal model. Extensive evaluations across 18 diverse benchmarks demonstrate its superior performance over strong open-source and leading proprietary frontier models.

Dong-Yang Li, Wang Zhao, Yuxin Chen et al.

Recent advances in 3D generative models have rapidly improved image-to-3D synthesis quality, enabling higher-resolution geometry and more realistic appearance. Yet fidelity, which measures pixel-level faithfulness of the generated 3D asset to the input image, still remains a central bottleneck. We argue this stems from an implicit 2D-3D correspondence issue: most 3D-native generators synthesize shape in canonical space and inject image cues via attention, leaving pixel-to-3D associations ambiguous. To tackle this issue, we draw inspiration from 3D reconstruction and propose Pixal3D, a pixel-aligned 3D generation paradigm for high-fidelity 3D asset creation from images. Instead of generating in a canonical pose, Pixal3D directly generates 3D in a pixel-aligned way, consistent with the input view. To enable this, we introduce a pixel back-projection conditioning scheme that explicitly lifts multi-scale image features into a 3D feature volume, establishing direct pixel-to-3D correspondence without ambiguity. We show that Pixal3D is not only scalable and capable of producing high-quality 3D assets, but also substantially improves fidelity, approaching the fidelity level of reconstruction. Furthermore, Pixal3D naturally extends to multi-view generation by aggregating back-projected feature volumes across views. Finally, we show pixel-aligned generation benefits scene synthesis, and present a modular pipeline that produces high-fidelity, object-separated 3D scenes from images. Pixal3D for the first time demonstrates 3D-native pixel-aligned generation at scale, and provides a new inspiring way towards high-fidelity 3D generation of object or scene from single or multi-view images. Project page: https://ldyang694.github.io/projects/pixal3d/

Yaowei Li, Xiaoyu Li, Zhaoyang Zhang et al.

Image customization, a crucial technique for industrial media production, aims to generate content that is consistent with reference images. However, current approaches conventionally separate image customization into position-aware and position-free customization paradigms and lack a universal framework for diverse customization, limiting their applications across various scenarios. To overcome these limitations, we propose IC-Custom, a unified framework that seamlessly integrates position-aware and position-free image customization through in-context learning. IC-Custom concatenates reference images with target images to a polyptych, leveraging DiT's multi-modal attention mechanism for fine-grained token-level interactions. We propose the In-context Multi-Modal Attention (ICMA) mechanism, which employs learnable task-oriented register tokens and boundary-aware positional embeddings to enable the model to effectively handle diverse tasks and distinguish between inputs in polyptych configurations. To address the data gap, we curated a 12K identity-consistent dataset with 8K real-world and 4K high-quality synthetic samples, avoiding the overly glossy, oversaturated look typical of synthetic data. IC-Custom supports various industrial applications, including try-on, image insertion, and creative IP customization. Extensive evaluations on our proposed ProductBench and the publicly available DreamBench demonstrate that IC-Custom significantly outperforms community workflows, closed-source models, and state-of-the-art open-source approaches. IC-Custom achieves about 73\% higher human preference across identity consistency, harmony, and text alignment metrics, while training only 0.4\% of the original model parameters. Project page: https://liyaowei-stu.github.io/project/IC_Custom

Cheng-Fu Yang, Da Yin, Wenbo Hu et al.

Humans recognize objects after observing only a few examples, a remarkable capability enabled by their inherent language understanding of the real-world environment. Developing verbalized and interpretable representation can significantly improve model generalization in low-data settings. In this work, we propose Verbalized Representation Learning (VRL), a novel approach for automatically extracting human-interpretable features for object recognition using few-shot data. Our method uniquely captures inter-class differences and intra-class commonalities in the form of natural language by employing a Vision-Language Model (VLM) to identify key discriminative features between different classes and shared characteristics within the same class. These verbalized features are then mapped to numeric vectors through the VLM. The resulting feature vectors can be further utilized to train and infer with downstream classifiers. Experimental results show that, at the same model scale, VRL achieves a 24% absolute improvement over prior state-of-the-art methods while using 95% less data and a smaller mode. Furthermore, compared to human-labeled attributes, the features learned by VRL exhibit a 20% absolute gain when used for downstream classification tasks. Code is available at: https://github.com/joeyy5588/VRL/tree/main.

Xinlei Xiong, Wenbo Hu, Shuxun Zhou et al.

Weather forecasting is fundamentally challenged by the chaotic nature of the atmosphere, necessitating probabilistic approaches to quantify uncertainty. While traditional ensemble prediction (EPS) addresses this through computationally intensive simulations, recent advances in Bayesian Deep Learning (BDL) offer a promising but often disconnected alternative. We bridge these paradigms through a unified hybrid Bayesian Deep Learning framework for ensemble weather forecasting that explicitly decomposes predictive uncertainty into epistemic and aleatoric components, learned via variational inference and a physics-informed stochastic perturbation scheme modeling flow-dependent atmospheric dynamics, respectively. We further establish a unified theoretical framework that rigorously connects BDL and EPS, providing formal theorems that decompose total predictive uncertainty into epistemic and aleatoric components under the hybrid BDL framework. We validate our framework on the large-scale 40-year ERA5 reanalysis dataset (1979-2019) with 0.25° spatial resolution. Experimental results show that our method not only improves forecast accuracy and yields better-calibrated uncertainty quantification but also achieves superior computational efficiency compared to state-of-the-art probabilistic diffusion models. We commit to making our code open-source upon acceptance of this paper.

Shuang Chen, Yue Guo, Yimeng Ye et al.

Recent advances in multimodal large reasoning models (MLRMs) have substantially improved their ability to solve complex textual and visual tasks. However, these models tend to overthink on simple problems, producing unnecessarily lengthy reasoning traces, while under-exploring on challenging ones, leading to missed solutions. To address this imbalance, we propose ARES, a unified open-source framework for adaptive reasoning that dynamically allocates exploration effort based on task difficulty. Our approach is motivated by two key empirical findings: (i) while single-token entropy is noisy, high window-entropy (HWE) tokens (token-level entropies averaged under a sliding window) can reliably capture reasoning-critical moments; and (ii) reducing HWE usage benefits easy problems, while increasing it is essential for solving hard ones. Building on these insights, ARES introduces a two-stage training pipeline. In the Adaptive Cold-Start stage, we curate multimodal and textual data paired with reasoning traces of length proportional to problem difficulty, equipping the model with initial difficulty awareness. In the second stage, we develop Adaptive Entropy Policy Optimization (AEPO), which uses HWE tokens as exploration triggers to decide when to explore, and a hierarchical entropy reward with dynamic KL control to decide how much to explore. Extensive experiments demonstrate that ARES achieves superior performance and reasoning efficiency across diverse mathematical, logical, and multimodal benchmarks, while closing the gap to leading commercial systems under significantly lower inference costs.

Junyi Zhang, Jia-Chen Gu, Wenbo Hu et al.

Existing medical reasoning benchmarks for vision-language models primarily focus on analyzing a patient's condition based on an image from a single visit. However, this setting deviates significantly from real-world clinical practice, where doctors typically refer to a patient's historical conditions to provide a comprehensive assessment by tracking their changes over time. In this paper, we introduce TemMed-Bench, the first benchmark designed for analyzing changes in patients' conditions between different clinical visits, which challenges large vision-language models (LVLMs) to reason over temporal medical images. TemMed-Bench consists of a test set comprising three tasks - visual question-answering (VQA), report generation, and image-pair selection - and a supplementary knowledge corpus of over 17,000 instances. With TemMed-Bench, we conduct an evaluation of six proprietary and six open-source LVLMs. Our results show that most LVLMs lack the ability to analyze patients' condition changes over temporal medical images, and a large proportion perform only at a random-guessing level in the closed-book setting. In contrast, GPT o3, o4-mini and Claude 3.5 Sonnet demonstrate comparatively decent performance, though they have yet to reach the desired level. Furthermore, we explore augmenting the input with both retrieved visual and textual modalities in the medical domain. We also show that multi-modal retrieval augmentation yields notably higher performance gains than no retrieval and textual retrieval alone across most models on our benchmark, with the VQA task showing an average improvement of 2.59%. Overall, we compose a benchmark grounded on real-world clinical practice, and it reveals LVLMs' limitations in temporal medical image reasoning, as well as highlighting the use of multi-modal retrieval augmentation as a potentially promising direction worth exploring to address this challenge.

Youze Wang, Zijun Chen, Ruoyu Chen et al.

Recent advancements in multimodal large language models for video understanding (videoLLMs) have enhanced their capacity to process complex spatiotemporal data. However, challenges such as factual inaccuracies, harmful content, biases, hallucinations, and privacy risks compromise their reliability. This study introduces Trust-videoLLMs, a first comprehensive benchmark evaluating 23 state-of-the-art videoLLMs (5 commercial, 18 open-source) across five critical dimensions: truthfulness, robustness, safety, fairness, and privacy. Comprising 30 tasks with adapted, synthetic, and annotated videos, the framework assesses spatiotemporal risks, temporal consistency and cross-modal impact. Results reveal significant limitations in dynamic scene comprehension, cross-modal perturbation resilience and real-world risk mitigation. While open-source models occasionally outperform, proprietary models generally exhibit superior credibility, though scaling does not consistently improve performance. These findings underscore the need for enhanced training datat diversity and robust multimodal alignment. Trust-videoLLMs provides a publicly available, extensible toolkit for standardized trustworthiness assessments, addressing the critical gap between accuracy-focused benchmarks and demands for robustness, safety, fairness, and privacy.

Jiaxin Gao, Wenbo Hu, Yuntian Chen

Long-term time series forecasting (LTSF) is a crucial aspect of modern society, playing a pivotal role in facilitating long-term planning and developing early warning systems. While many Transformer-based models have recently been introduced for LTSF, a doubt have been raised regarding the effectiveness of attention modules in capturing cross-time dependencies. In this study, we design a mask-series experiment to validate this assumption and subsequently propose the "Cross-variable Linear Integrated ENhanced Transformer for Multivariate Long-Term Time Series Forecasting" (Client), an advanced model that outperforms both traditional Transformer-based models and linear models. Client employs linear modules to learn trend information and attention modules to capture cross-variable dependencies. Meanwhile, it simplifies the embedding and position encoding layers and replaces the decoder module with a projection layer. Essentially, Client incorporates non-linearity and cross-variable dependencies, which sets it apart from conventional linear models and Transformer-based models. Extensive experiments with nine real-world datasets have confirmed the SOTA performance of Client with the least computation time and memory consumption compared with the previous Transformer-based models. Our code is available at https://github.com/daxin007/Client.

Wenbo Hu, Hengshuang Zhao, Li Jiang et al.

2D image representations are in regular grids and can be processed efficiently, whereas 3D point clouds are unordered and scattered in 3D space. The information inside these two visual domains is well complementary, e.g., 2D images have fine-grained texture while 3D point clouds contain plentiful geometry information. However, most current visual recognition systems process them individually. In this paper, we present a \emph{bidirectional projection network (BPNet)} for joint 2D and 3D reasoning in an end-to-end manner. It contains 2D and 3D sub-networks with symmetric architectures, that are connected by our proposed \emph{bidirectional projection module (BPM)}. Via the \emph{BPM}, complementary 2D and 3D information can interact with each other in multiple architectural levels, such that advantages in these two visual domains can be combined for better scene recognition. Extensive quantitative and qualitative experimental evaluations show that joint reasoning over 2D and 3D visual domains can benefit both 2D and 3D scene understanding simultaneously. Our \emph{BPNet} achieves top performance on the ScanNetV2 benchmark for both 2D and 3D semantic segmentation. Code is available at \url{https://github.com/wbhu/BPNet}.

Fangneng Zhan, Yingchen Yu, Changgong Zhang et al.

Inferring the scene illumination from a single image is an essential yet challenging task in computer vision and computer graphics. Existing works estimate lighting by regressing representative illumination parameters or generating illumination maps directly. However, these methods often suffer from poor accuracy and generalization. This paper presents Geometric Mover's Light (GMLight), a lighting estimation framework that employs a regression network and a generative projector for effective illumination estimation. We parameterize illumination scenes in terms of the geometric light distribution, light intensity, ambient term, and auxiliary depth, which can be estimated by a regression network. Inspired by the earth mover's distance, we design a novel geometric mover's loss to guide the accurate regression of light distribution parameters. With the estimated light parameters, the generative projector synthesizes panoramic illumination maps with realistic appearance and high-frequency details. Extensive experiments show that GMLight achieves accurate illumination estimation and superior fidelity in relighting for 3D object insertion. The codes are available at \href{https://github.com/fnzhan/Illumination-Estimation}{https://github.com/fnzhan/Illumination-Estimation}.

Zheng Zhang, Wenbo Hu, Yixing Lao et al.

3D Gaussian Splatting (3DGS) has demonstrated impressive novel view synthesis results while advancing real-time rendering performance. However, it relies heavily on the quality of the initial point cloud, resulting in blurring and needle-like artifacts in areas with insufficient initializing points. This is mainly attributed to the point cloud growth condition in 3DGS that only considers the average gradient magnitude of points from observable views, thereby failing to grow for large Gaussians that are observable for many viewpoints while many of them are only covered in the boundaries. To this end, we propose a novel method, named Pixel-GS, to take into account the number of pixels covered by the Gaussian in each view during the computation of the growth condition. We regard the covered pixel numbers as the weights to dynamically average the gradients from different views, such that the growth of large Gaussians can be prompted. As a result, points within the areas with insufficient initializing points can be grown more effectively, leading to a more accurate and detailed reconstruction. In addition, we propose a simple yet effective strategy to scale the gradient field according to the distance to the camera, to suppress the growth of floaters near the camera. Extensive experiments both qualitatively and quantitatively demonstrate that our method achieves state-of-the-art rendering quality while maintaining real-time rendering speed, on the challenging Mip-NeRF 360 and Tanks & Temples datasets.

Zhihao Liang, Qi Zhang, Wenbo Hu et al.

The 3D Gaussian Splatting (3DGS) gained its popularity recently by combining the advantages of both primitive-based and volumetric 3D representations, resulting in improved quality and efficiency for 3D scene rendering. However, 3DGS is not alias-free, and its rendering at varying resolutions could produce severe blurring or jaggies. This is because 3DGS treats each pixel as an isolated, single point rather than as an area, causing insensitivity to changes in the footprints of pixels. Consequently, this discrete sampling scheme inevitably results in aliasing, owing to the restricted sampling bandwidth. In this paper, we derive an analytical solution to address this issue. More specifically, we use a conditioned logistic function as the analytic approximation of the cumulative distribution function (CDF) in a one-dimensional Gaussian signal and calculate the Gaussian integral by subtracting the CDFs. We then introduce this approximation in the two-dimensional pixel shading, and present Analytic-Splatting, which analytically approximates the Gaussian integral within the 2D-pixel window area to better capture the intensity response of each pixel. Moreover, we use the approximated response of the pixel window integral area to participate in the transmittance calculation of volume rendering, making Analytic-Splatting sensitive to the changes in pixel footprint at different resolutions. Experiments on various datasets validate that our approach has better anti-aliasing capability that gives more details and better fidelity.

Mark YU, Wenbo Hu, Jinbo Xing et al.

We present TrajectoryCrafter, a novel approach to redirect camera trajectories for monocular videos. By disentangling deterministic view transformations from stochastic content generation, our method achieves precise control over user-specified camera trajectories. We propose a novel dual-stream conditional video diffusion model that concurrently integrates point cloud renders and source videos as conditions, ensuring accurate view transformations and coherent 4D content generation. Instead of leveraging scarce multi-view videos, we curate a hybrid training dataset combining web-scale monocular videos with static multi-view datasets, by our innovative double-reprojection strategy, significantly fostering robust generalization across diverse scenes. Extensive evaluations on multi-view and large-scale monocular videos demonstrate the superior performance of our method.

Tian-Xing Xu, Wenbo Hu, Yu-Kun Lai et al.

3D Gaussian splatting, emerging as a groundbreaking approach, has drawn increasing attention for its capabilities of high-fidelity reconstruction and real-time rendering. However, it couples the appearance and geometry of the scene within the Gaussian attributes, which hinders the flexibility of editing operations, such as texture swapping. To address this issue, we propose a novel approach, namely Texture-GS, to disentangle the appearance from the geometry by representing it as a 2D texture mapped onto the 3D surface, thereby facilitating appearance editing. Technically, the disentanglement is achieved by our proposed texture mapping module, which consists of a UV mapping MLP to learn the UV coordinates for the 3D Gaussian centers, a local Taylor expansion of the MLP to efficiently approximate the UV coordinates for the ray-Gaussian intersections, and a learnable texture to capture the fine-grained appearance. Extensive experiments on the DTU dataset demonstrate that our method not only facilitates high-fidelity appearance editing but also achieves real-time rendering on consumer-level devices, e.g. a single RTX 2080 Ti GPU.

Junchen Liu, Wenbo Hu, Zhuo Yang et al.

Despite significant advancements in Neural Radiance Fields (NeRFs), the renderings may still suffer from aliasing and blurring artifacts, since it remains a fundamental challenge to effectively and efficiently characterize anisotropic areas induced by the cone-casting procedure. This paper introduces a Ripmap-Encoded Platonic Solid representation to precisely and efficiently featurize 3D anisotropic areas, achieving high-fidelity anti-aliasing renderings. Central to our approach are two key components: Platonic Solid Projection and Ripmap encoding. The Platonic Solid Projection factorizes the 3D space onto the unparalleled faces of a certain Platonic solid, such that the anisotropic 3D areas can be projected onto planes with distinguishable characterization. Meanwhile, each face of the Platonic solid is encoded by the Ripmap encoding, which is constructed by anisotropically pre-filtering a learnable feature grid, to enable featurzing the projected anisotropic areas both precisely and efficiently by the anisotropic area-sampling. Extensive experiments on both well-established synthetic datasets and a newly captured real-world dataset demonstrate that our Rip-NeRF attains state-of-the-art rendering quality, particularly excelling in the fine details of repetitive structures and textures, while maintaining relatively swift training times.

Shuang Chen, Quanxin Shou, Hangting Chen et al.

Unified multimodal models provide a natural and promising architecture for understanding diverse and complex real-world knowledge while generating high-quality images. However, they still rely primarily on frozen parametric knowledge, which makes them struggle with real-world image generation involving long-tail and knowledge-intensive concepts. Inspired by the broad success of agents on real-world tasks, we explore agentic modeling to address this limitation. Specifically, we present Unify-Agent, a unified multimodal agent for world-grounded image synthesis, which reframes image generation as an agentic pipeline consisting of prompt understanding, multimodal evidence searching, grounded recaptioning, and final synthesis. To train our model, we construct a tailored multimodal data pipeline and curate 143K high-quality agent trajectories for world-grounded image synthesis, enabling effective supervision over the full agentic generation process. We further introduce FactIP, a benchmark covering 12 categories of culturally significant and long-tail factual concepts that explicitly requires external knowledge grounding. Extensive experiments show that our proposed Unify-Agent substantially improves over its base unified model across diverse benchmarks and real world generation tasks, while approaching the world knowledge capabilities of the strongest closed-source models. As an early exploration of agent-based modeling for world-grounded image synthesis, our work highlights the value of tightly coupling reasoning, searching, and generation for reliable open-world agentic image synthesis.

Haoyi Qiu, Wenbo Hu, Zi-Yi Dou et al.

Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs, undermining their reliability. A comprehensive quantitative evaluation is necessary to identify and understand the extent of hallucinations in these models. However, existing benchmarks are often limited in scope, focusing mainly on object hallucinations. Furthermore, current evaluation methods struggle to effectively address the subtle semantic distinctions between model outputs and reference data, as well as the balance between hallucination and informativeness. To address these issues, we introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases. Moreover, we propose a large language model (LLM)-based two-stage evaluation framework that generalizes the popular CHAIR metric and incorporates both faithfulness and coverage into the evaluation. Experiments on 10 established LVLMs demonstrate that our evaluation metric is more comprehensive and better correlated with humans than existing work when evaluating on our challenging human-annotated benchmark dataset. Our work also highlights the critical balance between faithfulness and coverage of model outputs, and encourages future works to address hallucinations in LVLMs while keeping their outputs informative.

Haoyuan Wang, Wenbo Hu, Lei Zhu et al.

Inverse rendering aims at recovering both geometry and materials of objects. It provides a more compatible reconstruction for conventional rendering engines, compared with the neural radiance fields (NeRFs). On the other hand, existing NeRF-based inverse rendering methods cannot handle glossy objects with local light interactions well, as they typically oversimplify the illumination as a 2D environmental map, which assumes infinite lights only. Observing the superiority of NeRFs in recovering radiance fields, we propose a novel 5D Neural Plenoptic Function (NeP) based on NeRFs and ray tracing, such that more accurate lighting-object interactions can be formulated via the rendering equation. We also design a material-aware cone sampling strategy to efficiently integrate lights inside the BRDF lobes with the help of pre-filtered radiance fields. Our method has two stages: the geometry of the target object and the pre-filtered environmental radiance fields are reconstructed in the first stage, and materials of the target object are estimated in the second stage with the proposed NeP and material-aware cone sampling strategy. Extensive experiments on the proposed real-world and synthetic datasets demonstrate that our method can reconstruct high-fidelity geometry/materials of challenging glossy objects with complex lighting interactions from nearby objects. Project webpage: https://whyy.site/paper/nep

Minghao Yin, Wenbo Hu, Jiale Xu et al.

Recent breakthroughs in 3D generative modeling have yielded remarkable progress in static shape synthesis, yet high-fidelity dynamic 4D generation remains elusive, hindered by temporal artifacts and prohibitive computational demand. We present Sculpt4D, a native 4D generative framework that seamlessly integrates efficient temporal modeling into a pretrained 3D Diffusion Transformer (Hunyuan3D 2.1), thereby mitigating the scarcity of 4D training data. At its core lies a Block Sparse Attention mechanism that preserves object identity by anchoring to the initial frame while capturing rich motion dynamics via a time-decaying sparse mask. This design faithfully models complex spatiotemporal dependencies with high fidelity, while sidestepping the quadratic overhead of full attention and reducing network total computation by 56%. Consequently, Sculpt4D establishes a new state-of-the-art in temporally coherent 4D synthesis and charts a path toward efficient and scalable 4D generation.

Tian-Xing Xu, Xiangjun Gao, Wenbo Hu et al.

Despite remarkable advancements in video depth estimation, existing methods exhibit inherent limitations in achieving geometric fidelity through the affine-invariant predictions, limiting their applicability in reconstruction and other metrically grounded downstream tasks. We propose GeometryCrafter, a novel framework that recovers high-fidelity point map sequences with temporal coherence from open-world videos, enabling accurate 3D/4D reconstruction, camera parameter estimation, and other depth-based applications. At the core of our approach lies a point map Variational Autoencoder (VAE) that learns a latent space agnostic to video latent distributions for effective point map encoding and decoding. Leveraging the VAE, we train a video diffusion model to model the distribution of point map sequences conditioned on the input videos. Extensive evaluations on diverse datasets demonstrate that GeometryCrafter achieves state-of-the-art 3D accuracy, temporal consistency, and generalization capability.

Xiaoning Dong, Wenbo Hu, Wei Xu et al.

Large language models (LLMs) have made significant advancements across various tasks, but their safety alignment remain a major concern. Exploring jailbreak prompts can expose LLMs' vulnerabilities and guide efforts to secure them. Existing methods primarily design sophisticated instructions for the LLM to follow, or rely on multiple iterations, which could hinder the performance and efficiency of jailbreaks. In this work, we propose a novel jailbreak paradigm, Simple Assistive Task Linkage (SATA), which can effectively circumvent LLM safeguards and elicit harmful responses. Specifically, SATA first masks harmful keywords within a malicious query to generate a relatively benign query containing one or multiple [MASK] special tokens. It then employs a simple assistive task such as a masked language model task or an element lookup by position task to encode the semantics of the masked keywords. Finally, SATA links the assistive task with the masked query to jointly perform the jailbreak. Extensive experiments show that SATA achieves state-of-the-art performance and outperforms baselines by a large margin. Specifically, on AdvBench dataset, with mask language model (MLM) assistive task, SATA achieves an overall attack success rate (ASR) of 85% and harmful score (HS) of 4.57, and with element lookup by position (ELP) assistive task, SATA attains an overall ASR of 76% and HS of 4.43.

Lingen Li, Zhaoyang Zhang, Yaowei Li et al.

Recent advancements in generative models have significantly improved novel view synthesis (NVS) from multi-view data. However, existing methods depend on external multi-view alignment processes, such as explicit pose estimation or pre-reconstruction, which limits their flexibility and accessibility, especially when alignment is unstable due to insufficient overlap or occlusions between views. In this paper, we propose NVComposer, a novel approach that eliminates the need for explicit external alignment. NVComposer enables the generative model to implicitly infer spatial and geometric relationships between multiple conditional views by introducing two key components: 1) an image-pose dual-stream diffusion model that simultaneously generates target novel views and condition camera poses, and 2) a geometry-aware feature alignment module that distills geometric priors from dense stereo models during training. Extensive experiments demonstrate that NVComposer achieves state-of-the-art performance in generative multi-view NVS tasks, removing the reliance on external alignment and thus improving model accessibility. Our approach shows substantial improvements in synthesis quality as the number of unposed input views increases, highlighting its potential for more flexible and accessible generative NVS systems. Our project page is available at https://lg-li.github.io/project/nvcomposer

Youze Wang, Wenbo Hu, Yinpeng Dong et al.

Large Language Models (LLMs) have evolved into Multimodal Large Language Models (MLLMs), significantly enhancing their capabilities by integrating visual information and other types, thus aligning more closely with the nature of human intelligence, which processes a variety of data forms beyond just text. Despite advancements, the undesirable generation of these models remains a critical concern, particularly due to vulnerabilities exposed by text-based jailbreak attacks, which have represented a significant threat by challenging existing safety protocols. Motivated by the unique security risks posed by the integration of new and old modalities for MLLMs, we propose a unified multimodal universal jailbreak attack framework that leverages iterative image-text interactions and transfer-based strategy to generate a universal adversarial suffix and image. Our work not only highlights the interaction of image-text modalities can be used as a critical vulnerability but also validates that multimodal universal jailbreak attacks can bring higher-quality undesirable generations across different MLLMs. We evaluate the undesirable context generation of MLLMs like LLaVA, Yi-VL, MiniGPT4, MiniGPT-v2, and InstructBLIP, and reveal significant multimodal safety alignment issues, highlighting the inadequacy of current safety mechanisms against sophisticated multimodal attacks. This study underscores the urgent need for robust safety measures in MLLMs, advocating for a comprehensive review and enhancement of security protocols to mitigate potential risks associated with multimodal capabilities.

Wenbo Hu, Yining Hong, Yanjun Wang et al.

Humans excel at performing complex tasks by leveraging long-term memory across temporal and spatial experiences. In contrast, current Large Language Models (LLMs) struggle to effectively plan and act in dynamic, multi-room 3D environments. We posit that part of this limitation is due to the lack of proper 3D spatial-temporal memory modeling in LLMs. To address this, we first introduce 3DMem-Bench, a comprehensive benchmark comprising over 26,000 trajectories and 2,892 embodied tasks, question-answering and captioning, designed to evaluate an agent's ability to reason over long-term memory in 3D environments. Second, we propose 3DLLM-Mem, a novel dynamic memory management and fusion model for embodied spatial-temporal reasoning and actions in LLMs. Our model uses working memory tokens, which represents current observations, as queries to selectively attend to and fuse the most useful spatial and temporal features from episodic memory, which stores past observations and interactions. Our approach allows the agent to focus on task-relevant information while maintaining memory efficiency in complex, long-horizon environments. Experimental results demonstrate that 3DLLM-Mem achieves state-of-the-art performance across various tasks, outperforming the strongest baselines by 16.5% in success rate on 3DMem-Bench's most challenging in-the-wild embodied tasks.

Jiaxin Gao, Wenbo Hu, Yuntian Chen

Revisiting PCA for Time Series Reduction in Temporal Dimension; Jiaxin Gao, Wenbo Hu, Yuntian Chen; Deep learning has significantly advanced time series analysis (TSA), enabling the extraction of complex patterns for tasks like classification, forecasting, and regression. Although dimensionality reduction has traditionally focused on the variable space-achieving notable success in minimizing data redundancy and computational complexity-less attention has been paid to reducing the temporal dimension. In this study, we revisit Principal Component Analysis (PCA), a classical dimensionality reduction technique, to explore its utility in temporal dimension reduction for time series data. It is generally thought that applying PCA to the temporal dimension would disrupt temporal dependencies, leading to limited exploration in this area. However, our theoretical analysis and extensive experiments demonstrate that applying PCA to sliding series windows not only maintains model performance, but also enhances computational efficiency. In auto-regressive forecasting, the temporal structure is partially preserved through windowing, and PCA is applied within these windows to denoise the time series while retaining their statistical information. By preprocessing time-series data with PCA, we reduce the temporal dimensionality before feeding it into TSA models such as Linear, Transformer, CNN, and RNN architectures. This approach accelerates training and inference and reduces resource consumption. Notably, PCA improves Informer training and inference speed by up to 40% and decreases GPU memory usage of TimesNet by 30%, without sacrificing model accuracy. Comparative analysis against other reduction methods further highlights the effectiveness of PCA in improving the efficiency of TSA models.

DaDong Jiang, Zhihui Ke, Xiaobo Zhou et al.

Dynamic scene reconstruction is a long-term challenge in 3D vision. Recent methods extend 3D Gaussian Splatting to dynamic scenes via additional deformation fields and apply explicit constraints like motion flow to guide the deformation. However, they learn motion changes from individual timestamps independently, making it challenging to reconstruct complex scenes, particularly when dealing with violent movement, extreme-shaped geometries, or reflective surfaces. To address the above issue, we design a plug-and-play module called TimeFormer to enable existing deformable 3D Gaussians reconstruction methods with the ability to implicitly model motion patterns from a learning perspective. Specifically, TimeFormer includes a Cross-Temporal Transformer Encoder, which adaptively learns the temporal relationships of deformable 3D Gaussians. Furthermore, we propose a two-stream optimization strategy that transfers the motion knowledge learned from TimeFormer to the base stream during the training phase. This allows us to remove TimeFormer during inference, thereby preserving the original rendering speed. Extensive experiments in the multi-view and monocular dynamic scenes validate qualitative and quantitative improvement brought by TimeFormer. Project Page: https://patrickddj.github.io/TimeFormer/

Kunhao Liu, Wenbo Hu, Jiale Xu et al.

Streaming video generation, as one fundamental component in interactive world models and neural game engines, aims to generate high-quality, low-latency, and temporally coherent long video streams. However, most existing work suffers from severe error accumulation that often significantly degrades the generated stream videos over long horizons. We design Rolling Forcing, a novel video generation technique that enables streaming long videos with minimal error accumulation. Rolling Forcing comes with three novel designs. First, instead of iteratively sampling individual frames, which accelerates error propagation, we design a joint denoising scheme that simultaneously denoises multiple frames with progressively increasing noise levels. This design relaxes the strict causality across adjacent frames, effectively suppressing error growth. Second, we introduce the attention sink mechanism into the long-horizon stream video generation task, which allows the model to keep key value states of initial frames as a global context anchor and thereby enhances long-term global consistency. Third, we design an efficient training algorithm that enables few-step distillation over largely extended denoising windows. This algorithm operates on non-overlapping windows and mitigates exposure bias conditioned on self-generated histories. Extensive experiments show that Rolling Forcing enables real-time streaming generation of multi-minute videos on a single GPU, with substantially reduced error accumulation.

Yanrui Bin, Wenbo Hu, Haoyuan Wang et al.

Surface normal estimation serves as a cornerstone for a spectrum of computer vision applications. While numerous efforts have been devoted to static image scenarios, ensuring temporal coherence in video-based normal estimation remains a formidable challenge. Instead of merely augmenting existing methods with temporal components, we present NormalCrafter to leverage the inherent temporal priors of video diffusion models. To secure high-fidelity normal estimation across sequences, we propose Semantic Feature Regularization (SFR), which aligns diffusion features with semantic cues, encouraging the model to concentrate on the intrinsic semantics of the scene. Moreover, we introduce a two-stage training protocol that leverages both latent and pixel space learning to preserve spatial accuracy while maintaining long temporal context. Extensive evaluations demonstrate the efficacy of our method, showcasing a superior performance in generating temporally consistent normal sequences with intricate details from diverse videos.

Yi Yao, Miao Fan, Shengtong Xu et al.

Lane topology reasoning techniques play a crucial role in high-definition (HD) mapping and autonomous driving applications. While recent years have witnessed significant advances in this field, there has been limited effort to consolidate these works into a comprehensive overview. This survey systematically reviews the evolution and current state of lane topology reasoning methods, categorizing them into three major paradigms: procedural modeling-based methods, aerial imagery-based methods, and onboard sensors-based methods. We analyze the progression from early rule-based approaches to modern learning-based solutions utilizing transformers, graph neural networks (GNNs), and other deep learning architectures. The paper examines standardized evaluation metrics, including road-level measures (APLS and TLTS score), and lane-level metrics (DET and TOP score), along with performance comparisons on benchmark datasets such as OpenLane-V2. We identify key technical challenges, including dataset availability and model efficiency, and outline promising directions for future research. This comprehensive review provides researchers and practitioners with insights into the theoretical frameworks, practical implementations, and emerging trends in lane topology reasoning for HD mapping applications.

Wenbo Hu, Jingli Lin, Yilin Long et al.

Vision-Language Models (VLMs) still lack robustness in spatial intelligence, demonstrating poor performance on spatial understanding and reasoning tasks. We attribute this gap to the absence of a visual geometry learning process capable of reconstructing 3D space from 2D images. We present G$^2$VLM, a geometry grounded vision-language model that bridges two fundamental aspects of spatial intelligence: spatial 3D reconstruction and spatial understanding. G$^2$VLM natively leverages learned 3D visual geometry features to directly predict 3D attributes and enhance spatial reasoning tasks via in-context learning and interleaved reasoning. Our unified design is highly scalable for spatial understanding: it trains on abundant multi-view image and video data, while simultaneously leveraging the benefits of 3D visual priors that are typically only derived from hard-to-collect annotations. Experimental results demonstrate G$^2$VLM is proficient in both tasks, achieving comparable results to state-of-the-art feed-forward 3D reconstruction models and achieving better or competitive results across spatial understanding and reasoning tasks. By unifying a semantically strong VLM with low-level 3D vision tasks, we hope G$^2$VLM can serve as a strong baseline for the community and unlock more future applications, such as 3D scene editing.

Zijun Chen, Wenbo Hu, Richang Hong

Chain of Thought (CoT) reasoning has demonstrated remarkable deep reasoning capabilities in both large language models (LLMs) and multimodal large language models (MLLMs). However, its reliability is often undermined by the accumulation of errors in intermediate steps. This paper introduces an novel approach to calibrate the CoT reasoning accuracy by leveraging the model's intrinsic veracity encoding. We discover that specific attention head activations reliably reflect the truthfulness of reasoning steps in CoT. Based on this insight, we train a confidence predictor to evaluate the correctness of each reasoning step using these truthfulness-sensitive activations, dynamically selecting the most plausible reasoning path via beam search. Experimental results demonstrate that our method significantly outperforms the state-of-the-art baselines (e.g., Few-Shot CoT, Self-Consistency, and Self-Evaluation Guided Beam Search) across the mathematical, symbolic, and commonsense reasoning tasks, exhibiting superior accuracy and reliability in both unimodal and multimodal settings. We further validate the approach on large reasoning models, confirming its applicability to specialized reasoning models. Additionally, we explore the role of the model's self-correction ability in CoT reasoning. This work provides a novel reliability improvement path for CoT reasoning with broad application potential.